- Email: [email protected]
Estimation of life expectancy, loss-of-life expectancy, and lifetime healthcare expenditures for schizophrenia in Taiwan
Schizophrenia Research 171 (2016) 97–102
Contents lists available at ScienceDirect
Schizophrenia Research journal homepage: www.elsevier.com/locate/schres
Estimation of life expectancy, loss-of-life expectancy, and lifetime healthcare expenditures for schizophrenia in Taiwan Chhian Hūi Lêng a,b, Ming Hui Chou c,d, Sheng-Hsiang Lin e, Yen Kuang Yang f, Jung-Der Wang c,g,⁎ a
Institute of Allied Health Sciences, College of Medicine, National Cheng Kung University, Tainan 701, Taiwan School of Medical Sociology and Social Work, Chung Shan Medical University, 402 Taichung, Taiwan Institute of Public Health, College of Medicine, National Cheng Kung University, Tainan 701, Taiwan d Psychiatric Hospital, Home of Philanthropy, Tainan 712, Taiwan e Institute of Clinical Medicine, College of Medicine, National Cheng Kung University, Tainan 701, Taiwan f Department of Psychiatry, National Cheng Kung University and Hospital, Tainan 701, Taiwan g Departments of Occupational and Environmental Medicine and Internal Medicine, National Cheng Kung University Hospital, Tainan 701, Taiwan b c
a r t i c l e
i n f o
Article history: Received 17 April 2015 Received in revised form 30 December 2015 Accepted 15 January 2016 Available online 23 January 2016 Keywords: Life expectancy Expected years of life lost Lifetime health care costs Schizophrenia Statistical method
a b s t r a c t By employing a novel semi-parametric extrapolation method, the life expectancies after the first hospitalization for schizophrenia and the associated lifetime healthcare expenditures were both estimated. Based on the linkage between the National Health Insurance Research Database and the National Mortality Registry of Taiwan, we have established a schizophrenic cohort for 2000–2010 and followed up to 2011. Survival function was estimated through Kaplan–Meier's method and extrapolated throughout life. We applied a simple linear regression to the logit-transformed survival ratio between the schizophrenic cohort and the sex-, age-matched referents via Monte Carlo simulation from the national life table. The monthly survival probability was multiplied by the average healthcare expenditures and summed throughout life to estimate the lifelong cost reimbursed by the National Health Insurance. The results showed that patients diagnosed at age 20–29 had the highest expected years of life lost (EYLL), 15 and 9 years, in men and women, respectively, with corresponding lifetime healthcare expenditures of USD 48,000 and 53,000. Males generally had higher health cost per life-year than their female counterparts across their lifespan. We applied the same method to the first 6 years of the cohort and extrapolated to 12 years, which showed that the relative biases for different age strata were less than 5%. We thus concluded that the semi-parametric extrapolation method might provide a timely estimation of lifetime outcomes for health care planning of schizophrenia. © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction Schizophrenia is, in many cases, chronic and disabling. Accelerating and validating the estimation of lifetime outcomes becomes timely to prioritize individual and national health actions. The shortened life expectancy (LE), also called the expected years of life lost (EYLL), of schizophrenia ranges from 7.8 to 22.5 years (Hannerz et al., 2001; Kodesh et al., 2012; Laursen, 2011; Lawrence et al., 2013; Nielsen et al., 2013; Nordentoft et al., 2013; Tiihonen et al., 2009) when compared with the general population. Most of the above studies applied life table methods on national prevalence data for years. However, life table methods usually require large, longitudinal follow-up cohorts to reflect a low incidence and chronic course of schizophrenia. Moreover, long-term outcomes estimated from incidence or treatment-/policy-
⁎ Corresponding author at: Institute of Public Health, College of Medicine, National Cheng Kung University, No. 1, University Road, Tainan City 701, Taiwan. E-mail address: [email protected] (J.-D. Wang).
naïve patients would be more accurate for showing trends and social impacts in comparison with those based on prevalent cohorts. Costs of schizophrenia are important for healthcare resources allocation. Despite few reports on the lifetime costs of schizophrenia, prevalence-based estimates are relatively popular. The yearly costs of psychotic disorders, mostly schizophrenia, ranked the third among 19 brain disorders in 30 European countries in 2010 (Olesen et al., 2012). Langley-Hawthorne (1997) calculated LE and lifetime costs of schizophrenia by utilizing Markov models with various transitional probabilities and related costs extracted from the literature. She projected a mean LE of 31 years and approximate lifetime cost of one million dollars (36% direct costs) per patient with schizophrenia in 1995 in Australia (Langley-Hawthorne, 1997). However, these estimates of lifetime costs are still waiting for corroboration. We have developed a semi-parametric method to estimate a lifelong survival function (Hwang and Wang, 1999), which can be multiplied with the average monthly cost of treating patients with schizophrenia to obtain lifetime costs. The method was mathematically verified to be valid, if constant excess hazard can be assumed (Fang et al., 2007),
http://dx.doi.org/10.1016/j.schres.2016.01.033 0920-9964/© 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
98
C.H. Lêng et al. / Schizophrenia Research 171 (2016) 97–102
schizophrenia and related comorbidities, were retrieved from the NHI database. Age- and gender-stratified annual incidence rates were calculated by dividing the number of new cases during the 11-year period with the total sum of the mid-year population of individual age- and gender-specific strata. 2.2. Extrapolation of long-term survival for the schizophrenic cohort
Fig. 1. (a) Logit transformation of the survival ratio W(t) between the survival functions of male patients with schizophrenia, aged 20–29, and that of the age- and gender-matched reference population generated by the Monte Carlo method. The two vertical dotted lines mark the time period when the logit survival ratio data were used for extrapolation. The bottom dotted line is the linear regression line. (b) Lifelong extrapolation to estimate the life expectancy of the 20–29 schizophrenia cohort and the corresponding reference.
and was empirically demonstrated to estimate life expectancy (LE), and by extension expected years of life lost (EYLL) on cancer cohorts with high censored rates (Andersson et al., 2013; Liu et al., 2013). Recently, it was also applied on cohorts of heroin users with relative accuracy (Chang et al., 2015). This study aims to estimate the LE, EYLL, lifetime healthcare expenditures, and cost-per-life year for a cohort of incident cases of schizophrenia followed for 12 years in Taiwan. 2. Materials and methods 2.1. Schizophrenia cohort The National Health Insurance (NHI) in Taiwan was implemented in 1995; and since 2004, more than 99% of Taiwan citizens have been covered (National Health Insurance, 2005). The catastrophic illness system of the NHI features an up-to-date list of severe illnesses, including schizophrenia, and the system waives the co-payment for the treatment of the listed diseases. Schizophrenia healthcare that follows the formal guidelines in Taiwan is freely accessible to registered patients diagnosed and validated with code 295 based on ICD-9-CM by at least two boardcertified psychiatrists. We identified a total of 58,665 patients with schizophrenia first registered with the NHI as catastrophic illness over the period of 2000–2010. Among them, 38,429 patients were hospitalized for schizophrenia for the first time. In the final analysis, only 34,658 within the age range of 20–64 when registered were included to determine if the assumption of constant excess hazard could be fulfilled among various age groups. The enrollees were either censored at death or at the end of 2011. Their survival status was verified by linkage with the national mortality registry. The health care expenditures, including the inpatient and outpatient expenditures attributed to
We applied Kaplan–Meier's method to estimate the time-to-mortality of the schizophrenic cohort stratified by gender and age. Lifetime survival function was estimated by a semi-parametric extrapolation method developed by Hwang and Wang (Hwang et al., 1996; Hwang and Wang, 1999; Hwang and Wang, 2004) and later mathematically validated by Fang et al. (2007). The method is briefly summarized as follows: first, hypothetical age- and gender-specific referents for each patient with schizophrenia were simulated by the Monte Carlo method conditioned on the hazard function of the life table in a given year in Taiwan. Second, the survival ratio between schizophrenic patients and referents at time t was calculated, and assumed to be one initially and gradually converge to a constant, representing the excess hazard of the premature mortality accompanying schizophrenia. Third, the survival ratio was then logit-transformed so that a simple linear regression could be applied to approximate the temporal trend of the stabilized period until the end of 2011. Accordingly, the slope of the regression line was used to extrapolate the lifetime survival of schizophrenia patients after the follow-up limit, allowing the life expectancy (LE) of an age-, gender-specific cohort to be estimated by summing up the area under the curve of the survival function throughout the patient's life. The bootstrap method of repeated sampling for 100 iterations was used to compute the mean standard error. Fig. 1(a) illustrates the slope of constant excess hazard for extrapolation (a) in male patients with schizophrenia, aged 20–29. Shaded area of the projected survival curve indicates the EYLL of 20–29 male patients (Fig. 1(b)). A software package developed by Hwang's team was used for the computation, and can be freely downloaded: Integration of Survival with Quality of Life. http://www. stat.sinica.edu.tw/isqol/ (updated on Feb. 12, 2014). 2.3. Estimation of expected years of life lost and lifetime healthcare expenditures for schizophrenia We estimated two lifetime outcomes to assess the impact of schizophrenia. First, the expected years of life lost (EYLL) was the difference in LE between the schizophrenic cohort and the age- and gender-matched hypothetical referents, as described above. The second indicator was the average lifetime healthcare expenditure per person, for which the reimbursement data of the National Health Insurance from 2000 to 2010 included all medical expenses attributed to schizophrenia. We stratified the cohort by 10-year age and gender specific strata, and then calculated
Table 1 Annual incidence rate per 100,000, age at diagnosis and gender of cohorts. Schizophrenia (ICD9: 295)
Gender
Total number of new cases during 2000–2010
Age at diagnosis, mean (SD)
Annual incidence rate (×100,000−1)
20–29
Male Female Male Female Male Female Male Female Male Female Male Female
6372 4197 6333 4645 3972 3908 1774 2431 402 624
24.72(2.89) 24.85(2.84) 34.30(2.86) 34.42(2.89) 43.97(2.86) 44.27(2.87) 53.69(2.76) 53.77(2.81) 61.88(1.44) 61.73(1.36)
30.21 20.73 30.47 22.77 19.52 19.50 12.53 16.99 9.23 13.59 968.7 864.1
30–39 40–49 50–59 60–64 Cumulative incidence rates (CIR20–64)
C.H. Lêng et al. / Schizophrenia Research 171 (2016) 97–102
99
Table 2 Life expectancy, expected years of life lost, and health care expenditures (USD) of schizophrenia stratified by sex and age. Schizophrenia (ICD9: 295)
Gender
Number
Life expectancy Mean (SE) years
Expected years of life lost Mean (SE) years
20–29
Male Female Male Female Male Female Male Female Male Female
6372 4197 6333 4645 3972 3908 1774 2431 402 624
35.64(0.03) 47.17(0.04) 34.74(0.04) 40.99(0.05) 26.77(0.04) 29.48(0.05) 18.4(0.08) 21.34(0.05) 14.44(0.2) 17.77(0.14)
15.17(0.04) 9.22(0.03) 7.28(0.03) 6.19(0.04) 6.66(0.04) 8.24(0.04) 6.9(0.08) 7.66(0.06) 4.50(0.20) 4.21(0.14)
30–39 40–49 50–59 60–64 a b
Life time healthcare expenditure/case (3% discount)a 48,244 ± 3079 (42,741–54,557) 53,411 ± 2905 (48,359–58,911) 43,205 ± 2867 (38,628–49,179) 46,260 ± 3097 (42,214–53,673) 44,423 ± 2542 (40,112–50,528) 44,229 ± 3009 (39,927–51,206) 38,791 ± 2010 (36,320–43,755) 38,214 ± 1793 (34,793–41,024) 34,279 ± 2130 (30,527–38,431) 37,568 ± 1434 (35,136–40,864)
Cost per life year (PPP-adjusted)b 4061.73 (260.71) 3886.94 (212.56) 3754.46 (249.79) 3604.02 (242.19) 4530.65 (262.41) 4184.75 (286.25) 5126.53 (259.77) 4468.32 (209.34) 5536.98 (348.23) 5110.21 (211.74)
1USD$ = 29.322 New Taiwan Dollars (NTD) on December 31, 2010. Purchasing Power Parities (PPP) adjusted exchange rate: 1USD = 16.83 NTD (2010, https://www.imf.org).
the average monthly cost by dividing the total healthcare expenditures of all survived patients with the number of survived patients at that month, of which the expenditure was first adjusted to the nominal exchange rate of 1USD = 29.322 New Taiwan Dollars (NTD) on December 31, 2010. Then, we multiplied the average monthly expenditure by the corresponding survival probability and summed up over a lifetime to obtain lifelong cost. The extrapolated costs were adjusted with an annual discount rate of 3%. We also obtained the cost per life year by dividing the lifetime expenditure with the discounted LE, 3% per year, for consistency. To be more comparable internationally, the cost per life-year was adjusted according to the purchasing power parities (PPP) implied exchange rate: 1USD = 16.83 NTD, (International Monetary Fund. http://www.imf.org) in 2010.
in Table 1. Males first registering their schizophrenia as a catastrophic illness peaked at a younger age than for females. 3.1. Estimation of life expectancy (LE) and expected years of life lost (EYLL) Stratified by age and gender, Table 2 summarizes the LE, EYLL, lifetime healthcare costs, and cost per life year. On average, males with schizophrenia had higher EYLL's than did females, except for the 40– 59 age group. Males aged 20–29 had the highest EYLL, 15.17 years, and dropped to approximately 6–7 years if onset at the age of 30–59. Female patients aged 20–29 had the first EYLL peak of 9.2 years, while the second EYLL peak of 7–8 years occurred for the 40–59 year-old females. 3.2. Lifetime healthcare expenditures and cost per life year
2.4. Validation of the extrapolation method We validated our extrapolation method as follows: The first 6-year follow-up data (registered during 2000–2005) was extrapolated for 6 years to estimate the 12-year LE, which were compared with the Kaplan–Meier estimate based on the actual 12-year follow-up to the end of 2011. Namely, out-of-sample predictive validity tests were based on comparing predictions for the remaining 6 years of data withheld from the model-building exercise with the actual observed data. In addition, every graph of logit transformed survival ratio was plotted and examined if the assumption of constant excess hazard was fulfilled near the end of 12-year follow-up. 3. Results The 34,658 patients with schizophrenia were stratified by gender and age. The total number of subjects, mean of diagnosis age, and annual incidence rate of each age- and gender-specific group are summarized
Table 2 illustrates the estimated life time healthcare expenditures with 3% discount (USD), and PPP-adjusted cost per life year with 3% discount on both costs. The highest lifetime expenditures were $48,244 ± 3079 for males aged 20–29; this figure descended to a plateau of approximately $43,000–$44,000 before age 50, and then gradually decreased as the registering age increased. Despite the similar LE of 35 years for both the 20–29 and 30–39 male groups, the younger group cost $5039 more in lifetime health expenditures than did the 30–39 group. This might imply a difference in the illness severity in the younger onset groups despite a similar chronicity. By contrast, female patients with schizophrenia who registered at 20–39 had longer life expectancies and higher overall lifetime expenditures than their male counterparts. Females who registered at the age of 20–29 had the highest lifetime costs, $53,411 ± 2905, among all gender and age groups. After age 40, their lifetime expenditures were similar to those of the same age male groups and also decreased according to age. Furthermore, health expenditure per life year ranged from USD$3600 to
Table 3 Validation of the extrapolated estimates according to 6 years of follow-up and the Kaplan–Meier method. Schizophrenia (ICD9: 295)
Gender
Cohort size
Registered age (SD)
Censored rate (%)
12(2000–2011)-year follow-up Kaplan–Meier estimate Mean (SE) months
Estimate using the extrapolation based on the first 6 years of follow-up Mean (SE) months
Relative biasa %
20–29
Male Female Male Female Male Female Male Female Male Female
4385 2795 4297 2825 2628 2503 1005 1470 245 411
24.58 (2.89) 24.74 (2.86) 34.35 (2.84) 34.47 (2.90) 43.90 (2.83) 44.20 (2.88) 53.54 (2.76) 53.61 (2.83) 62.02 (1.43) 61.75 (1.35)
98.02 98.10 96.02 97.52 94.71 96.56 92.24 95.71 86.12 91.73
11.50 (0.03) 11.54 (0.04) 11.08 (0.04) 11.44 (0.04) 10.76 (0.06) 11.18 (0.05) 10.00 (0.12) 10.79 (0.07) 8.64 (0.27) 9.80 (0.16)
11.53 (0.01) 11.62 (0.01) 11.26 (0.01) 11.43 (0.01) 10.79 (0.03) 11.30 (0.02) 10.42 (0.05) 11.20 (0.02) 8.14 (0.13) 9.54 (0.06)
0.30 0.66 1.57 −0.06 0.32 1.03 4.18 3.82 −5.81 −2.63
30–39 40–49 50–59 60–64 a
Relative bias = (estimate from extrapolation – K–M estimate) / K–M estimate.
100
C.H. Lêng et al. / Schizophrenia Research 171 (2016) 97–102
$5550 (PPP-adjusted). Moreover, healthcare expenditures per person year generally increased as the subjects aged. 3.3. Validation of the semi-parametric method Table 3 indicates the differences between the semi-parametric extrapolation of 6 years based on the first 6 years and the observed 12year LE. The relative biases were less than 5% in all gender-age groups, with the exception of males aged 60–64, who had a 5.81% relative bias, an absolute difference of 0.5 year, and the lowest censored rate (86.12%). Notably, as the sample size decreased, the relative bias seemed less stable. Fig. 2 illustrates the constant excess hazard of each stratified age and gender stratum. The assumption of constant excess
hazard appears to be satisfied among the patients aged 20–64 because the logit transformation of the survival ratio between patients with schizophrenia and the age- and gender-matched referents shows a linear trend and estimable slope approximately 1–2 years after diagnosis. 4. Discussion This study reveals that onset of schizophrenia appears to be associated with substantial EYLL, ranging from 6 to 15 years, while associated lifetime healthcare expenditures range from USD 38,000 to 53,000 per person (Table 2). Both estimates depend heavily on accurate estimation of the lifetime survival function of the schizophrenic cohort. Thus, we
Fig. 2. The survival rate ratio W(t) between the patients with schizophrenia and that of the age- and gender-matched referents was transformed into logit (W(t) / (1 − W(t))), which is plotted against time after diagnosis. It converges to a straight line for all the 10 different age and gender strata after the initial period of 1–2 years, indicating constant excess hazard exists and the assumption is fulfilled. The two dotted lines mark the time period when the slope of logit of survival ratio was used for extrapolation.
C.H. Lêng et al. / Schizophrenia Research 171 (2016) 97–102
should first verify the estimation validity of our lifelong survival function before making any further inferences. Onset of schizophrenia cascades a series of acute and chronic physical and psychosocial difficulties. As the course stabilizes, evaluation of long-term outcomes might require extrapolation beyond the limited follow-up time. To consider both the excess hazard associated with schizophrenia and the background hazard of the general population, we graphed the logittransformed relative survival and fit a regression line along the temporally-stabilized period near the end of 12-year of follow-up for extrapolation (Figs. 1(a) and 2). The relative bias estimated by our semi-parametric extrapolation seems robust when compared with the Kaplan–Meier estimates after 12 years of real follow-up (Table 3). To control the cohort effect due to competing risks of mortality or longterm side effects of medication, we used incidence cases and stratified them by age and gender. Our analysis shows that cohorts of schizophrenia generally fulfill the assumption of constant excess hazard, namely, there is an estimable slope of logit of survival ratio (W(t)) after the first several years, which have lasted up to 12 years (Figs. 1(a) and 2). Since we have stratified our cohorts into every 10-year age period, we would expect that such a constant excess mortality would lead to the next 10-year age periods and our extrapolation would most likely be valid. Accordingly, the estimated lifetime health expenditures paid by the NHI of Taiwan would also be relatively accurate. The sex ratio and incidence rates of schizophrenia in Taiwan (Table 1) seem similar to those reported from other countries (McGrath et al., 2004). The EYLL were found to be 15.17 and 9.22 years for male and female, respectively, for schizophrenia aged 20–29 (Table 2). As our cohort did not include schizophrenia diagnosed below age 20, who are expected to have an even higher EYLL, we believed that the overall estimates would be similar to those reported by several western countries, for example, 14.6 in southeast London (Chang et al., 2011), 15.3–20.1 in Denmark (Laursen, 2011; Nielsen et al., 2013; Nordentoft et al., 2013), 12.8–22.5 in Finland (Nordentoft et al., 2013; Tiihonen et al., 2009; Westman et al., 2012), and 16.4 in Western Australia (Lawrence et al., 2013). The EYLL for age over 30 appear to be more stabilized and smaller than that of the younger age group, which corroborates a previous study (Chen et al., 1996). Because Taiwan is a small island with 36,000 km2 and has implemented a universal coverage healthcare system since 1995, residents can generally access healthcare services within two hours of transportation. The above unique condition may have lowered down mortality of acute care (Wen et al., 2008), including suicide or infections (Nordentoft et al., 2013), and possibly schizophrenia related illnesses (Laursen, 2011; Nielsen et al., 2013). Nevertheless, future studies of stratifying different causes of mortality are warranted to corroborate the above hypothesis. Patients with schizophrenia onset before age 50, had relatively high lifetime healthcare costs, ranging from USD 43,000 to 49,000 per person (Table 2). The lifetime costs would be underestimated for some patients, however, because the NHI restricted the hospitalized patients to no more than 6 months and some patients would be transferred to nursing homes and switch to the use of social welfare resources for a short period of time. The reimbursement from most psychiatric nursing homes was from the Department of Social Affairs, Ministry of the Interior before 2012, and is not included in the present study. But the reimbursement of medication is still paid by the NHI. Furthermore, divided by discounted LE, healthcare costs per life year ranged from USD 3600 to 5500 (PPP-adjusted), while the corresponding costs ranged from €5102–7068 (€PPP, 2010) in France, Germany, Spain, and the UK (Gustavsson et al., 2011). Adjusting for PPP decreased the disparity. In addition to yearly estimation, our approximated lifelong costs could bolster a comprehensive long-term care or budget plans. This study has at least the following limitations. First, the registered age, however, is used as a proxy of the onset for all patients, which would result in underestimation of EYLL because of possible registry delays. Since all registered patients with schizophrenia are waived from all
101
copayments, the financial incentive generally reduces such a potential bias to a minimum. Second, we have not systematically collected any data related to out-of-pocket costs (Lee et al., 2008). As patients with schizophrenia usually receive numerous alternative and complimentary treatments in traditional Taiwanese culture, such a discrepancy should not be ignored and future studies must include this portion when tallying social costs. In conclusion, we have successfully applied a novel semi-parametric extrapolation to estimate gender- and age- specific LE, EYLL, and the lifetime healthcare expenditures for schizophrenia in Taiwan. The strategy of this method is based on follow-up for more than 10–12 years and stratification of a cohort into every 10-year age strata to examine if constant excess mortality exists in each age strata. And it could also be applied in other serious mental illnesses leading to premature mortality, e.g., bipolar disorder and/or major depression. Schizophrenia first diagnosed at age 20–29 had the highest EYLL, 15 and 9 years, as well as the largest lifetime healthcare expenditures, USD 48,000 and 53,000, in men and women, respectively. Future studies are indicated for longer followup terms and data collection of out-of-pocket money to determine the entire burden to society. Funding body agreements and policies The study was supported by the National Science Council, Taiwan, NSC 101-3114-Y-006-001. The funding source had no role in study design, analysis, interpretation, and any other process in the paper. Contributors All authors designed the study. JDW performed the analysis. CHL prepared the first draft of the manuscript under the supervision of JDW. All authors commented on the manuscript and have approved the final version. Conflict of interest All authors declare that they have no conflict of interest. Acknowledgments We have been keeping in mind the suffering and efforts of many psychotic patients and their families.
References Andersson, T.M., Dickman, P.W., Eloranta, S., Lambe, M., Lambert, P.C., 2013. Estimating the loss in expectation of life due to cancer using flexible parametric survival models. Stat. Med. 32 (30), 5286–5300. Chang, C.K., Hayes, R.D., Perera, G., Broadbent, M.T., Fernandes, A.C., Lee, W.E., Hotopf, M., Stewart, R., 2011. Life expectancy at birth for people with serious mental illness and other major disorders from a secondary mental health care case register in London. PLoS One 6 (5), e19590. Chang, K.C., Lu, C.H., Hwang, J.S., Wang, J.D., 2015. Estimation of life expectancy and the expected years of life lost among heroin users in the era of opioid substitution treatment (OST) in Taiwan. Drug Alcohol Depend. 153, 152–158. Chen, W.J., Huang, Y.J., Yeh, L.L., Rin, H., Hwu, H.G., 1996. Excess mortality of psychiatric inpatients in Taiwan. Psychiatry Res. 62 (3), 239–250. Fang, C.T., Chang, Y.Y., Hsu, H.M., Twu, S.J., Chen, K.T., Lin, C.C., Huang, L.Y., Chen, M.Y., Hwang, J.S., Wang, J.D., Chuang, C.Y., 2007. Life expectancy of patients with newlydiagnosed HIV infection in the era of highly active antiretroviral therapy. QJM 100 (2), 97–105. Gustavsson, A., Svensson, M., Jacobi, F., Allgulander, C., Alonso, J., Beghi, E., Dodel, R., Ekman, M., Faravelli, C., Fratiglioni, L., Gannon, B., Jones, D.H., Jennum, P., Jordanova, A., Jonsson, L., Karampampa, K., Knapp, M., Kobelt, G., Kurth, T., Lieb, R., Linde, M., Ljungcrantz, C., Maercker, A., Melin, B., Moscarelli, M., Musayev, A., Norwood, F., Preisig, M., Pugliatti, M., Rehm, J., Salvador-Carulla, L., Schlehofer, B., Simon, R., Steinhausen, H.C., Stovner, L.J., Vallat, J.M., Van den Bergh, P., van Os, J., Vos, P., Xu, W., Wittchen, H.U., Jonsson, B., Olesen, J., Group, C.D., 2011. Cost of disorders of the brain in Europe 2010. Eur. Neuropsychopharmacol. 21 (10), 718–779. Hannerz, H., Borga, P., Borritz, M., 2001. Life expectancies for individuals with psychiatric diagnoses. Public Health 115 (5), 328–337. Hwang, J.S., Wang, J.D., 1999. Monte Carlo estimation of extrapolation of quality-adjusted survival for follow-up studies. Stat. Med. 18 (13), 1627–1640. Hwang, J.S., Wang, J.D., 2004. Integrating health profile with survival for quality of life assessment. Qual. Life Res. 13 (1), 1–10 discussion 11-14. Hwang, J.S., Tsauo, J.Y., Wang, J.D., 1996. Estimation of expected quality adjusted survival by cross-sectional survey. Stat. Med. 15 (1), 93–102. Integration of Survival with Quality of Life. http://www.stat.sinica.edu.tw/isqol/ (last accessed Dec. 16, 2015). International Monetary Fund. http://www.imf.org (last accessed Apr. 10, 2015).
102
C.H. Lêng et al. / Schizophrenia Research 171 (2016) 97–102
Kodesh, A., Goldshtein, I., Gelkopf, M., Goren, I., Chodick, G., Shalev, V., 2012. Epidemiology and comorbidity of severe mental illnesses in the community: findings from a computerized mental health registry in a large Israeli health organization. Soc. Psychiatry Psychiatr. Epidemiol. 47 (11), 1775–1782. Langley-Hawthorne, C., 1997. Modeling the lifetime costs of treating schizophrenia in Australia. Clin. Ther. 19 (6), 1470–1495 discussion 1424-1475. Laursen, T.M., 2011. Life expectancy among persons with schizophrenia or bipolar affective disorder. Schizophr. Res. 131 (1–3), 101–104. Lawrence, D., Hancock, K.J., Kisely, S., 2013. The gap in life expectancy from preventable physical illness in psychiatric patients in Western Australia: retrospective analysis of population based registers. BMJ 346, f2539. Lee, I.H., Chen, P.S., Yang, Y.K., Liao, Y.C., Lee, Y.D., Yeh, T.L., Yeh, L.L., Cheng, S.H., Chu, C.L., 2008. The functionality and economic costs of outpatients with schizophrenia in Taiwan. Psychiatry Res. 158 (3), 306–315. Liu, P.H., Wang, J.D., Keating, N.L., 2013. Expected years of life lost for six potentially preventable cancers in the United States. Prev. Med. 56 (5), 309–313. McGrath, J., Saha, S., Welham, J., El Saadi, O., MacCauley, C., Chant, D., 2004. A systematic review of the incidence of schizophrenia: the distribution of rates and the influence of sex, urbanicity, migrant status and methodology. BMC Med. 2, 13. National Health Insurance, 2005. http://www.nhi.gov.tw/English/webdata/webdata. aspx?menu=11&menu_id=296&webdata_id=1942&WD_ID=296 (last accessed Dec. 28, 2015).
Nielsen, R.E., Uggerby, A.S., Jensen, S.O., McGrath, J.J., 2013. Increasing mortality gap for patients diagnosed with schizophrenia over the last three decades—a Danish nationwide study from 1980 to 2010. Schizophr. Res. 146 (1–3), 22–27. Nordentoft, M., Wahlbeck, K., Hallgren, J., Westman, J., Osby, U., Alinaghizadeh, H., Gissler, M., Laursen, T.M., 2013. Excess mortality, causes of death and life expectancy in 270,770 patients with recent onset of mental disorders in Denmark, Finland and Sweden. PLoS One 8 (1), e55176. Olesen, J., Gustavsson, A., Svensson, M., Wittchen, H.U., Jonsson, B., Group, C.S., European Brain, C., 2012. The economic cost of brain disorders in Europe. Eur. J. Neurol. 19 (1), 155–162. Tiihonen, J., Lonnqvist, J., Wahlbeck, K., Klaukka, T., Niskanen, L., Tanskanen, A., Haukka, J., 2009. 11-year follow-up of mortality in patients with schizophrenia: a populationbased cohort study (FIN11 study). Lancet 374 (9690), 620–627. Wen, C.P., Tsai, S.P., Chung, W.S., 2008. A 10-year experience with universal health insurance in Taiwan: measuring changes in health and health disparity. Ann. Intern. Med. 148 (4), 258–267. Westman, J., Gissler, M., Wahlbeck, K., 2012. Successful deinstitutionalization of mental health care: increased life expectancy among people with mental disorders in Finland. Eur. J. Pub. Health 22 (4), 604–606.
Contents lists available at ScienceDirect
Schizophrenia Research journal homepage: www.elsevier.com/locate/schres
Estimation of life expectancy, loss-of-life expectancy, and lifetime healthcare expenditures for schizophrenia in Taiwan Chhian Hūi Lêng a,b, Ming Hui Chou c,d, Sheng-Hsiang Lin e, Yen Kuang Yang f, Jung-Der Wang c,g,⁎ a
Institute of Allied Health Sciences, College of Medicine, National Cheng Kung University, Tainan 701, Taiwan School of Medical Sociology and Social Work, Chung Shan Medical University, 402 Taichung, Taiwan Institute of Public Health, College of Medicine, National Cheng Kung University, Tainan 701, Taiwan d Psychiatric Hospital, Home of Philanthropy, Tainan 712, Taiwan e Institute of Clinical Medicine, College of Medicine, National Cheng Kung University, Tainan 701, Taiwan f Department of Psychiatry, National Cheng Kung University and Hospital, Tainan 701, Taiwan g Departments of Occupational and Environmental Medicine and Internal Medicine, National Cheng Kung University Hospital, Tainan 701, Taiwan b c
a r t i c l e
i n f o
Article history: Received 17 April 2015 Received in revised form 30 December 2015 Accepted 15 January 2016 Available online 23 January 2016 Keywords: Life expectancy Expected years of life lost Lifetime health care costs Schizophrenia Statistical method
a b s t r a c t By employing a novel semi-parametric extrapolation method, the life expectancies after the first hospitalization for schizophrenia and the associated lifetime healthcare expenditures were both estimated. Based on the linkage between the National Health Insurance Research Database and the National Mortality Registry of Taiwan, we have established a schizophrenic cohort for 2000–2010 and followed up to 2011. Survival function was estimated through Kaplan–Meier's method and extrapolated throughout life. We applied a simple linear regression to the logit-transformed survival ratio between the schizophrenic cohort and the sex-, age-matched referents via Monte Carlo simulation from the national life table. The monthly survival probability was multiplied by the average healthcare expenditures and summed throughout life to estimate the lifelong cost reimbursed by the National Health Insurance. The results showed that patients diagnosed at age 20–29 had the highest expected years of life lost (EYLL), 15 and 9 years, in men and women, respectively, with corresponding lifetime healthcare expenditures of USD 48,000 and 53,000. Males generally had higher health cost per life-year than their female counterparts across their lifespan. We applied the same method to the first 6 years of the cohort and extrapolated to 12 years, which showed that the relative biases for different age strata were less than 5%. We thus concluded that the semi-parametric extrapolation method might provide a timely estimation of lifetime outcomes for health care planning of schizophrenia. © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction Schizophrenia is, in many cases, chronic and disabling. Accelerating and validating the estimation of lifetime outcomes becomes timely to prioritize individual and national health actions. The shortened life expectancy (LE), also called the expected years of life lost (EYLL), of schizophrenia ranges from 7.8 to 22.5 years (Hannerz et al., 2001; Kodesh et al., 2012; Laursen, 2011; Lawrence et al., 2013; Nielsen et al., 2013; Nordentoft et al., 2013; Tiihonen et al., 2009) when compared with the general population. Most of the above studies applied life table methods on national prevalence data for years. However, life table methods usually require large, longitudinal follow-up cohorts to reflect a low incidence and chronic course of schizophrenia. Moreover, long-term outcomes estimated from incidence or treatment-/policy-
⁎ Corresponding author at: Institute of Public Health, College of Medicine, National Cheng Kung University, No. 1, University Road, Tainan City 701, Taiwan. E-mail address: [email protected] (J.-D. Wang).
naïve patients would be more accurate for showing trends and social impacts in comparison with those based on prevalent cohorts. Costs of schizophrenia are important for healthcare resources allocation. Despite few reports on the lifetime costs of schizophrenia, prevalence-based estimates are relatively popular. The yearly costs of psychotic disorders, mostly schizophrenia, ranked the third among 19 brain disorders in 30 European countries in 2010 (Olesen et al., 2012). Langley-Hawthorne (1997) calculated LE and lifetime costs of schizophrenia by utilizing Markov models with various transitional probabilities and related costs extracted from the literature. She projected a mean LE of 31 years and approximate lifetime cost of one million dollars (36% direct costs) per patient with schizophrenia in 1995 in Australia (Langley-Hawthorne, 1997). However, these estimates of lifetime costs are still waiting for corroboration. We have developed a semi-parametric method to estimate a lifelong survival function (Hwang and Wang, 1999), which can be multiplied with the average monthly cost of treating patients with schizophrenia to obtain lifetime costs. The method was mathematically verified to be valid, if constant excess hazard can be assumed (Fang et al., 2007),
http://dx.doi.org/10.1016/j.schres.2016.01.033 0920-9964/© 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
98
C.H. Lêng et al. / Schizophrenia Research 171 (2016) 97–102
schizophrenia and related comorbidities, were retrieved from the NHI database. Age- and gender-stratified annual incidence rates were calculated by dividing the number of new cases during the 11-year period with the total sum of the mid-year population of individual age- and gender-specific strata. 2.2. Extrapolation of long-term survival for the schizophrenic cohort
Fig. 1. (a) Logit transformation of the survival ratio W(t) between the survival functions of male patients with schizophrenia, aged 20–29, and that of the age- and gender-matched reference population generated by the Monte Carlo method. The two vertical dotted lines mark the time period when the logit survival ratio data were used for extrapolation. The bottom dotted line is the linear regression line. (b) Lifelong extrapolation to estimate the life expectancy of the 20–29 schizophrenia cohort and the corresponding reference.
and was empirically demonstrated to estimate life expectancy (LE), and by extension expected years of life lost (EYLL) on cancer cohorts with high censored rates (Andersson et al., 2013; Liu et al., 2013). Recently, it was also applied on cohorts of heroin users with relative accuracy (Chang et al., 2015). This study aims to estimate the LE, EYLL, lifetime healthcare expenditures, and cost-per-life year for a cohort of incident cases of schizophrenia followed for 12 years in Taiwan. 2. Materials and methods 2.1. Schizophrenia cohort The National Health Insurance (NHI) in Taiwan was implemented in 1995; and since 2004, more than 99% of Taiwan citizens have been covered (National Health Insurance, 2005). The catastrophic illness system of the NHI features an up-to-date list of severe illnesses, including schizophrenia, and the system waives the co-payment for the treatment of the listed diseases. Schizophrenia healthcare that follows the formal guidelines in Taiwan is freely accessible to registered patients diagnosed and validated with code 295 based on ICD-9-CM by at least two boardcertified psychiatrists. We identified a total of 58,665 patients with schizophrenia first registered with the NHI as catastrophic illness over the period of 2000–2010. Among them, 38,429 patients were hospitalized for schizophrenia for the first time. In the final analysis, only 34,658 within the age range of 20–64 when registered were included to determine if the assumption of constant excess hazard could be fulfilled among various age groups. The enrollees were either censored at death or at the end of 2011. Their survival status was verified by linkage with the national mortality registry. The health care expenditures, including the inpatient and outpatient expenditures attributed to
We applied Kaplan–Meier's method to estimate the time-to-mortality of the schizophrenic cohort stratified by gender and age. Lifetime survival function was estimated by a semi-parametric extrapolation method developed by Hwang and Wang (Hwang et al., 1996; Hwang and Wang, 1999; Hwang and Wang, 2004) and later mathematically validated by Fang et al. (2007). The method is briefly summarized as follows: first, hypothetical age- and gender-specific referents for each patient with schizophrenia were simulated by the Monte Carlo method conditioned on the hazard function of the life table in a given year in Taiwan. Second, the survival ratio between schizophrenic patients and referents at time t was calculated, and assumed to be one initially and gradually converge to a constant, representing the excess hazard of the premature mortality accompanying schizophrenia. Third, the survival ratio was then logit-transformed so that a simple linear regression could be applied to approximate the temporal trend of the stabilized period until the end of 2011. Accordingly, the slope of the regression line was used to extrapolate the lifetime survival of schizophrenia patients after the follow-up limit, allowing the life expectancy (LE) of an age-, gender-specific cohort to be estimated by summing up the area under the curve of the survival function throughout the patient's life. The bootstrap method of repeated sampling for 100 iterations was used to compute the mean standard error. Fig. 1(a) illustrates the slope of constant excess hazard for extrapolation (a) in male patients with schizophrenia, aged 20–29. Shaded area of the projected survival curve indicates the EYLL of 20–29 male patients (Fig. 1(b)). A software package developed by Hwang's team was used for the computation, and can be freely downloaded: Integration of Survival with Quality of Life. http://www. stat.sinica.edu.tw/isqol/ (updated on Feb. 12, 2014). 2.3. Estimation of expected years of life lost and lifetime healthcare expenditures for schizophrenia We estimated two lifetime outcomes to assess the impact of schizophrenia. First, the expected years of life lost (EYLL) was the difference in LE between the schizophrenic cohort and the age- and gender-matched hypothetical referents, as described above. The second indicator was the average lifetime healthcare expenditure per person, for which the reimbursement data of the National Health Insurance from 2000 to 2010 included all medical expenses attributed to schizophrenia. We stratified the cohort by 10-year age and gender specific strata, and then calculated
Table 1 Annual incidence rate per 100,000, age at diagnosis and gender of cohorts. Schizophrenia (ICD9: 295)
Gender
Total number of new cases during 2000–2010
Age at diagnosis, mean (SD)
Annual incidence rate (×100,000−1)
20–29
Male Female Male Female Male Female Male Female Male Female Male Female
6372 4197 6333 4645 3972 3908 1774 2431 402 624
24.72(2.89) 24.85(2.84) 34.30(2.86) 34.42(2.89) 43.97(2.86) 44.27(2.87) 53.69(2.76) 53.77(2.81) 61.88(1.44) 61.73(1.36)
30.21 20.73 30.47 22.77 19.52 19.50 12.53 16.99 9.23 13.59 968.7 864.1
30–39 40–49 50–59 60–64 Cumulative incidence rates (CIR20–64)
C.H. Lêng et al. / Schizophrenia Research 171 (2016) 97–102
99
Table 2 Life expectancy, expected years of life lost, and health care expenditures (USD) of schizophrenia stratified by sex and age. Schizophrenia (ICD9: 295)
Gender
Number
Life expectancy Mean (SE) years
Expected years of life lost Mean (SE) years
20–29
Male Female Male Female Male Female Male Female Male Female
6372 4197 6333 4645 3972 3908 1774 2431 402 624
35.64(0.03) 47.17(0.04) 34.74(0.04) 40.99(0.05) 26.77(0.04) 29.48(0.05) 18.4(0.08) 21.34(0.05) 14.44(0.2) 17.77(0.14)
15.17(0.04) 9.22(0.03) 7.28(0.03) 6.19(0.04) 6.66(0.04) 8.24(0.04) 6.9(0.08) 7.66(0.06) 4.50(0.20) 4.21(0.14)
30–39 40–49 50–59 60–64 a b
Life time healthcare expenditure/case (3% discount)a 48,244 ± 3079 (42,741–54,557) 53,411 ± 2905 (48,359–58,911) 43,205 ± 2867 (38,628–49,179) 46,260 ± 3097 (42,214–53,673) 44,423 ± 2542 (40,112–50,528) 44,229 ± 3009 (39,927–51,206) 38,791 ± 2010 (36,320–43,755) 38,214 ± 1793 (34,793–41,024) 34,279 ± 2130 (30,527–38,431) 37,568 ± 1434 (35,136–40,864)
Cost per life year (PPP-adjusted)b 4061.73 (260.71) 3886.94 (212.56) 3754.46 (249.79) 3604.02 (242.19) 4530.65 (262.41) 4184.75 (286.25) 5126.53 (259.77) 4468.32 (209.34) 5536.98 (348.23) 5110.21 (211.74)
1USD$ = 29.322 New Taiwan Dollars (NTD) on December 31, 2010. Purchasing Power Parities (PPP) adjusted exchange rate: 1USD = 16.83 NTD (2010, https://www.imf.org).
the average monthly cost by dividing the total healthcare expenditures of all survived patients with the number of survived patients at that month, of which the expenditure was first adjusted to the nominal exchange rate of 1USD = 29.322 New Taiwan Dollars (NTD) on December 31, 2010. Then, we multiplied the average monthly expenditure by the corresponding survival probability and summed up over a lifetime to obtain lifelong cost. The extrapolated costs were adjusted with an annual discount rate of 3%. We also obtained the cost per life year by dividing the lifetime expenditure with the discounted LE, 3% per year, for consistency. To be more comparable internationally, the cost per life-year was adjusted according to the purchasing power parities (PPP) implied exchange rate: 1USD = 16.83 NTD, (International Monetary Fund. http://www.imf.org) in 2010.
in Table 1. Males first registering their schizophrenia as a catastrophic illness peaked at a younger age than for females. 3.1. Estimation of life expectancy (LE) and expected years of life lost (EYLL) Stratified by age and gender, Table 2 summarizes the LE, EYLL, lifetime healthcare costs, and cost per life year. On average, males with schizophrenia had higher EYLL's than did females, except for the 40– 59 age group. Males aged 20–29 had the highest EYLL, 15.17 years, and dropped to approximately 6–7 years if onset at the age of 30–59. Female patients aged 20–29 had the first EYLL peak of 9.2 years, while the second EYLL peak of 7–8 years occurred for the 40–59 year-old females. 3.2. Lifetime healthcare expenditures and cost per life year
2.4. Validation of the extrapolation method We validated our extrapolation method as follows: The first 6-year follow-up data (registered during 2000–2005) was extrapolated for 6 years to estimate the 12-year LE, which were compared with the Kaplan–Meier estimate based on the actual 12-year follow-up to the end of 2011. Namely, out-of-sample predictive validity tests were based on comparing predictions for the remaining 6 years of data withheld from the model-building exercise with the actual observed data. In addition, every graph of logit transformed survival ratio was plotted and examined if the assumption of constant excess hazard was fulfilled near the end of 12-year follow-up. 3. Results The 34,658 patients with schizophrenia were stratified by gender and age. The total number of subjects, mean of diagnosis age, and annual incidence rate of each age- and gender-specific group are summarized
Table 2 illustrates the estimated life time healthcare expenditures with 3% discount (USD), and PPP-adjusted cost per life year with 3% discount on both costs. The highest lifetime expenditures were $48,244 ± 3079 for males aged 20–29; this figure descended to a plateau of approximately $43,000–$44,000 before age 50, and then gradually decreased as the registering age increased. Despite the similar LE of 35 years for both the 20–29 and 30–39 male groups, the younger group cost $5039 more in lifetime health expenditures than did the 30–39 group. This might imply a difference in the illness severity in the younger onset groups despite a similar chronicity. By contrast, female patients with schizophrenia who registered at 20–39 had longer life expectancies and higher overall lifetime expenditures than their male counterparts. Females who registered at the age of 20–29 had the highest lifetime costs, $53,411 ± 2905, among all gender and age groups. After age 40, their lifetime expenditures were similar to those of the same age male groups and also decreased according to age. Furthermore, health expenditure per life year ranged from USD$3600 to
Table 3 Validation of the extrapolated estimates according to 6 years of follow-up and the Kaplan–Meier method. Schizophrenia (ICD9: 295)
Gender
Cohort size
Registered age (SD)
Censored rate (%)
12(2000–2011)-year follow-up Kaplan–Meier estimate Mean (SE) months
Estimate using the extrapolation based on the first 6 years of follow-up Mean (SE) months
Relative biasa %
20–29
Male Female Male Female Male Female Male Female Male Female
4385 2795 4297 2825 2628 2503 1005 1470 245 411
24.58 (2.89) 24.74 (2.86) 34.35 (2.84) 34.47 (2.90) 43.90 (2.83) 44.20 (2.88) 53.54 (2.76) 53.61 (2.83) 62.02 (1.43) 61.75 (1.35)
98.02 98.10 96.02 97.52 94.71 96.56 92.24 95.71 86.12 91.73
11.50 (0.03) 11.54 (0.04) 11.08 (0.04) 11.44 (0.04) 10.76 (0.06) 11.18 (0.05) 10.00 (0.12) 10.79 (0.07) 8.64 (0.27) 9.80 (0.16)
11.53 (0.01) 11.62 (0.01) 11.26 (0.01) 11.43 (0.01) 10.79 (0.03) 11.30 (0.02) 10.42 (0.05) 11.20 (0.02) 8.14 (0.13) 9.54 (0.06)
0.30 0.66 1.57 −0.06 0.32 1.03 4.18 3.82 −5.81 −2.63
30–39 40–49 50–59 60–64 a
Relative bias = (estimate from extrapolation – K–M estimate) / K–M estimate.
100
C.H. Lêng et al. / Schizophrenia Research 171 (2016) 97–102
$5550 (PPP-adjusted). Moreover, healthcare expenditures per person year generally increased as the subjects aged. 3.3. Validation of the semi-parametric method Table 3 indicates the differences between the semi-parametric extrapolation of 6 years based on the first 6 years and the observed 12year LE. The relative biases were less than 5% in all gender-age groups, with the exception of males aged 60–64, who had a 5.81% relative bias, an absolute difference of 0.5 year, and the lowest censored rate (86.12%). Notably, as the sample size decreased, the relative bias seemed less stable. Fig. 2 illustrates the constant excess hazard of each stratified age and gender stratum. The assumption of constant excess
hazard appears to be satisfied among the patients aged 20–64 because the logit transformation of the survival ratio between patients with schizophrenia and the age- and gender-matched referents shows a linear trend and estimable slope approximately 1–2 years after diagnosis. 4. Discussion This study reveals that onset of schizophrenia appears to be associated with substantial EYLL, ranging from 6 to 15 years, while associated lifetime healthcare expenditures range from USD 38,000 to 53,000 per person (Table 2). Both estimates depend heavily on accurate estimation of the lifetime survival function of the schizophrenic cohort. Thus, we
Fig. 2. The survival rate ratio W(t) between the patients with schizophrenia and that of the age- and gender-matched referents was transformed into logit (W(t) / (1 − W(t))), which is plotted against time after diagnosis. It converges to a straight line for all the 10 different age and gender strata after the initial period of 1–2 years, indicating constant excess hazard exists and the assumption is fulfilled. The two dotted lines mark the time period when the slope of logit of survival ratio was used for extrapolation.
C.H. Lêng et al. / Schizophrenia Research 171 (2016) 97–102
should first verify the estimation validity of our lifelong survival function before making any further inferences. Onset of schizophrenia cascades a series of acute and chronic physical and psychosocial difficulties. As the course stabilizes, evaluation of long-term outcomes might require extrapolation beyond the limited follow-up time. To consider both the excess hazard associated with schizophrenia and the background hazard of the general population, we graphed the logittransformed relative survival and fit a regression line along the temporally-stabilized period near the end of 12-year of follow-up for extrapolation (Figs. 1(a) and 2). The relative bias estimated by our semi-parametric extrapolation seems robust when compared with the Kaplan–Meier estimates after 12 years of real follow-up (Table 3). To control the cohort effect due to competing risks of mortality or longterm side effects of medication, we used incidence cases and stratified them by age and gender. Our analysis shows that cohorts of schizophrenia generally fulfill the assumption of constant excess hazard, namely, there is an estimable slope of logit of survival ratio (W(t)) after the first several years, which have lasted up to 12 years (Figs. 1(a) and 2). Since we have stratified our cohorts into every 10-year age period, we would expect that such a constant excess mortality would lead to the next 10-year age periods and our extrapolation would most likely be valid. Accordingly, the estimated lifetime health expenditures paid by the NHI of Taiwan would also be relatively accurate. The sex ratio and incidence rates of schizophrenia in Taiwan (Table 1) seem similar to those reported from other countries (McGrath et al., 2004). The EYLL were found to be 15.17 and 9.22 years for male and female, respectively, for schizophrenia aged 20–29 (Table 2). As our cohort did not include schizophrenia diagnosed below age 20, who are expected to have an even higher EYLL, we believed that the overall estimates would be similar to those reported by several western countries, for example, 14.6 in southeast London (Chang et al., 2011), 15.3–20.1 in Denmark (Laursen, 2011; Nielsen et al., 2013; Nordentoft et al., 2013), 12.8–22.5 in Finland (Nordentoft et al., 2013; Tiihonen et al., 2009; Westman et al., 2012), and 16.4 in Western Australia (Lawrence et al., 2013). The EYLL for age over 30 appear to be more stabilized and smaller than that of the younger age group, which corroborates a previous study (Chen et al., 1996). Because Taiwan is a small island with 36,000 km2 and has implemented a universal coverage healthcare system since 1995, residents can generally access healthcare services within two hours of transportation. The above unique condition may have lowered down mortality of acute care (Wen et al., 2008), including suicide or infections (Nordentoft et al., 2013), and possibly schizophrenia related illnesses (Laursen, 2011; Nielsen et al., 2013). Nevertheless, future studies of stratifying different causes of mortality are warranted to corroborate the above hypothesis. Patients with schizophrenia onset before age 50, had relatively high lifetime healthcare costs, ranging from USD 43,000 to 49,000 per person (Table 2). The lifetime costs would be underestimated for some patients, however, because the NHI restricted the hospitalized patients to no more than 6 months and some patients would be transferred to nursing homes and switch to the use of social welfare resources for a short period of time. The reimbursement from most psychiatric nursing homes was from the Department of Social Affairs, Ministry of the Interior before 2012, and is not included in the present study. But the reimbursement of medication is still paid by the NHI. Furthermore, divided by discounted LE, healthcare costs per life year ranged from USD 3600 to 5500 (PPP-adjusted), while the corresponding costs ranged from €5102–7068 (€PPP, 2010) in France, Germany, Spain, and the UK (Gustavsson et al., 2011). Adjusting for PPP decreased the disparity. In addition to yearly estimation, our approximated lifelong costs could bolster a comprehensive long-term care or budget plans. This study has at least the following limitations. First, the registered age, however, is used as a proxy of the onset for all patients, which would result in underestimation of EYLL because of possible registry delays. Since all registered patients with schizophrenia are waived from all
101
copayments, the financial incentive generally reduces such a potential bias to a minimum. Second, we have not systematically collected any data related to out-of-pocket costs (Lee et al., 2008). As patients with schizophrenia usually receive numerous alternative and complimentary treatments in traditional Taiwanese culture, such a discrepancy should not be ignored and future studies must include this portion when tallying social costs. In conclusion, we have successfully applied a novel semi-parametric extrapolation to estimate gender- and age- specific LE, EYLL, and the lifetime healthcare expenditures for schizophrenia in Taiwan. The strategy of this method is based on follow-up for more than 10–12 years and stratification of a cohort into every 10-year age strata to examine if constant excess mortality exists in each age strata. And it could also be applied in other serious mental illnesses leading to premature mortality, e.g., bipolar disorder and/or major depression. Schizophrenia first diagnosed at age 20–29 had the highest EYLL, 15 and 9 years, as well as the largest lifetime healthcare expenditures, USD 48,000 and 53,000, in men and women, respectively. Future studies are indicated for longer followup terms and data collection of out-of-pocket money to determine the entire burden to society. Funding body agreements and policies The study was supported by the National Science Council, Taiwan, NSC 101-3114-Y-006-001. The funding source had no role in study design, analysis, interpretation, and any other process in the paper. Contributors All authors designed the study. JDW performed the analysis. CHL prepared the first draft of the manuscript under the supervision of JDW. All authors commented on the manuscript and have approved the final version. Conflict of interest All authors declare that they have no conflict of interest. Acknowledgments We have been keeping in mind the suffering and efforts of many psychotic patients and their families.
References Andersson, T.M., Dickman, P.W., Eloranta, S., Lambe, M., Lambert, P.C., 2013. Estimating the loss in expectation of life due to cancer using flexible parametric survival models. Stat. Med. 32 (30), 5286–5300. Chang, C.K., Hayes, R.D., Perera, G., Broadbent, M.T., Fernandes, A.C., Lee, W.E., Hotopf, M., Stewart, R., 2011. Life expectancy at birth for people with serious mental illness and other major disorders from a secondary mental health care case register in London. PLoS One 6 (5), e19590. Chang, K.C., Lu, C.H., Hwang, J.S., Wang, J.D., 2015. Estimation of life expectancy and the expected years of life lost among heroin users in the era of opioid substitution treatment (OST) in Taiwan. Drug Alcohol Depend. 153, 152–158. Chen, W.J., Huang, Y.J., Yeh, L.L., Rin, H., Hwu, H.G., 1996. Excess mortality of psychiatric inpatients in Taiwan. Psychiatry Res. 62 (3), 239–250. Fang, C.T., Chang, Y.Y., Hsu, H.M., Twu, S.J., Chen, K.T., Lin, C.C., Huang, L.Y., Chen, M.Y., Hwang, J.S., Wang, J.D., Chuang, C.Y., 2007. Life expectancy of patients with newlydiagnosed HIV infection in the era of highly active antiretroviral therapy. QJM 100 (2), 97–105. Gustavsson, A., Svensson, M., Jacobi, F., Allgulander, C., Alonso, J., Beghi, E., Dodel, R., Ekman, M., Faravelli, C., Fratiglioni, L., Gannon, B., Jones, D.H., Jennum, P., Jordanova, A., Jonsson, L., Karampampa, K., Knapp, M., Kobelt, G., Kurth, T., Lieb, R., Linde, M., Ljungcrantz, C., Maercker, A., Melin, B., Moscarelli, M., Musayev, A., Norwood, F., Preisig, M., Pugliatti, M., Rehm, J., Salvador-Carulla, L., Schlehofer, B., Simon, R., Steinhausen, H.C., Stovner, L.J., Vallat, J.M., Van den Bergh, P., van Os, J., Vos, P., Xu, W., Wittchen, H.U., Jonsson, B., Olesen, J., Group, C.D., 2011. Cost of disorders of the brain in Europe 2010. Eur. Neuropsychopharmacol. 21 (10), 718–779. Hannerz, H., Borga, P., Borritz, M., 2001. Life expectancies for individuals with psychiatric diagnoses. Public Health 115 (5), 328–337. Hwang, J.S., Wang, J.D., 1999. Monte Carlo estimation of extrapolation of quality-adjusted survival for follow-up studies. Stat. Med. 18 (13), 1627–1640. Hwang, J.S., Wang, J.D., 2004. Integrating health profile with survival for quality of life assessment. Qual. Life Res. 13 (1), 1–10 discussion 11-14. Hwang, J.S., Tsauo, J.Y., Wang, J.D., 1996. Estimation of expected quality adjusted survival by cross-sectional survey. Stat. Med. 15 (1), 93–102. Integration of Survival with Quality of Life. http://www.stat.sinica.edu.tw/isqol/ (last accessed Dec. 16, 2015). International Monetary Fund. http://www.imf.org (last accessed Apr. 10, 2015).
102
C.H. Lêng et al. / Schizophrenia Research 171 (2016) 97–102
Kodesh, A., Goldshtein, I., Gelkopf, M., Goren, I., Chodick, G., Shalev, V., 2012. Epidemiology and comorbidity of severe mental illnesses in the community: findings from a computerized mental health registry in a large Israeli health organization. Soc. Psychiatry Psychiatr. Epidemiol. 47 (11), 1775–1782. Langley-Hawthorne, C., 1997. Modeling the lifetime costs of treating schizophrenia in Australia. Clin. Ther. 19 (6), 1470–1495 discussion 1424-1475. Laursen, T.M., 2011. Life expectancy among persons with schizophrenia or bipolar affective disorder. Schizophr. Res. 131 (1–3), 101–104. Lawrence, D., Hancock, K.J., Kisely, S., 2013. The gap in life expectancy from preventable physical illness in psychiatric patients in Western Australia: retrospective analysis of population based registers. BMJ 346, f2539. Lee, I.H., Chen, P.S., Yang, Y.K., Liao, Y.C., Lee, Y.D., Yeh, T.L., Yeh, L.L., Cheng, S.H., Chu, C.L., 2008. The functionality and economic costs of outpatients with schizophrenia in Taiwan. Psychiatry Res. 158 (3), 306–315. Liu, P.H., Wang, J.D., Keating, N.L., 2013. Expected years of life lost for six potentially preventable cancers in the United States. Prev. Med. 56 (5), 309–313. McGrath, J., Saha, S., Welham, J., El Saadi, O., MacCauley, C., Chant, D., 2004. A systematic review of the incidence of schizophrenia: the distribution of rates and the influence of sex, urbanicity, migrant status and methodology. BMC Med. 2, 13. National Health Insurance, 2005. http://www.nhi.gov.tw/English/webdata/webdata. aspx?menu=11&menu_id=296&webdata_id=1942&WD_ID=296 (last accessed Dec. 28, 2015).
Nielsen, R.E., Uggerby, A.S., Jensen, S.O., McGrath, J.J., 2013. Increasing mortality gap for patients diagnosed with schizophrenia over the last three decades—a Danish nationwide study from 1980 to 2010. Schizophr. Res. 146 (1–3), 22–27. Nordentoft, M., Wahlbeck, K., Hallgren, J., Westman, J., Osby, U., Alinaghizadeh, H., Gissler, M., Laursen, T.M., 2013. Excess mortality, causes of death and life expectancy in 270,770 patients with recent onset of mental disorders in Denmark, Finland and Sweden. PLoS One 8 (1), e55176. Olesen, J., Gustavsson, A., Svensson, M., Wittchen, H.U., Jonsson, B., Group, C.S., European Brain, C., 2012. The economic cost of brain disorders in Europe. Eur. J. Neurol. 19 (1), 155–162. Tiihonen, J., Lonnqvist, J., Wahlbeck, K., Klaukka, T., Niskanen, L., Tanskanen, A., Haukka, J., 2009. 11-year follow-up of mortality in patients with schizophrenia: a populationbased cohort study (FIN11 study). Lancet 374 (9690), 620–627. Wen, C.P., Tsai, S.P., Chung, W.S., 2008. A 10-year experience with universal health insurance in Taiwan: measuring changes in health and health disparity. Ann. Intern. Med. 148 (4), 258–267. Westman, J., Gissler, M., Wahlbeck, K., 2012. Successful deinstitutionalization of mental health care: increased life expectancy among people with mental disorders in Finland. Eur. J. Pub. Health 22 (4), 604–606.